Present and future of machine learning
Ever wondered how your social media apps can get you glued to the screen? How do pharmaceutical companies come up with groundbreaking new drugs? How resources are increasingly becoming cheaper (except for the inflation periods in the macro economy)? Chances are it's a Machine Learning (ML) algorithm that is working behind the scenes.
Since it is already recorded I can add little value by providing an exact transcript in this article. But I will share what is in it with some details so you can listen to the podcast should it contain areas that interest you. A lot of care has been given to make sure anyone (both tech and non-tech) can understand what is discussed. So don't feel intimidated by technical-sounding words.
What is ML, and why it is needed?
If you Google search 'what is ML' you would get a definition something like 'ML is a subset of Artificial Intelligence (AI) that can use data to learn'. In the podcast, we took it much further than that.
We started by exploring what is ML, and why it is needed. As the definition says it is a subset of AI. We, humans, are great at learning and applying things. Traditional computers/software are great at processing large amounts of data. If you give a spreadsheet to a human with a million values we would not understand anything but any computer can quickly understand it and create a chart from it.
Similarly, if you give a traditional computer/software an image of a bird it cannot understand it because it cannot learn from past images of birds to know that this is a bird. Whereas even a child can easily identify the bird.
Machine learning brings advanced capabilities to traditional computers. Machine learning can not only process the data like a spreadsheet package but also can learn from new interactions.
An example would be, Google Maps for instance can not only keep track of where you have visited and know you can routinely visit a particular place at a particular time but also can adapt to changes. It can also understand your, say type of restaurants you like and suggest new restaurants.
Real-world applications - globally
Machine learning is everywhere. It would be impossible to cover everything like a laundry list. But to structure them, I see machine learning being used in three predominant ways. To improve efficiencies & asset usage (Eg: Google using ML to use data centre electricity efficiently), to improve healthcare (Eg: Pfizer using ML in creating drugs) and for our day-to-day work (Eg: Siri using ML to suggest apps).
Real-world applications - #SriLanka
The usage of machine learning in Sri Lanka is not as published as that of developed countries. So I had to explore with friends what case studies they know about machine learning being used in Sri Lanka. We can categorise the things I learned into three areas: performance improvements (heavily used in the apparel sector), quantity predictions (retail sector) and predictive maintenance (both consumer and enterprise electronics).
Will AI/ML replace the jobs?
Since it is a hot topic to discuss if AI/ML will replace the jobs in the future we discussed this. I do not believe AI/ML will replace humans. Like I said in the introduction, although computers can now learn still humans play an important role in the decision-making process. This is the same view taken by icons such as Peter Thiel (PayPal Co-founder) and Garry Kasparov (Chess World Champion). While routine jobs will be largely automated there will also be new job titles that do not exist in the present day. The best way to prepare for this future is to upskill yourself. If you look at openings from even traditional investment banks like JPMorgan Chase or Goldman Sachs, even they heavily hire tech-related talent.
How to learn AI/ML
How can someone (both tech and non-tech) get started? There are two ways to get started. You can develop your own custom solutions by coding from scratch or using existing machine learning packages in providers like Microsoft Azure, Amazon SageMaker, Apple ML Kit, etc. I self-taught myself out of curiosity and eventually published a research paper on my final year research [link] (a machine learning solution). So, I also covered some of my personal experiences so you can validate some of your experiences if you relate to the journey I took.
Using technology to empower developing countries
For the Sri Lankan audience, we also discussed a bit how technology can help Sri Lanka to overcome the economic challenges - something other developing countries could adopt too. We talked about how technology relatively has fewer barriers to entry. So, both individuals and governments can use this to their advantage to upskill themselves and thereby use the skill to export to high-demand countries. Essentially using technology to build their countries like China did with their manufacturing sector.
